Observability with Prometheus and Grafana

Using the TSDB Status Information

These tables are instrumental in diagnosing and mitigating performance issues related to high cardinality and memory usage.

By identifying the top 10 label names with the highest count of unique values, administrators can pinpoint labels that unexpectedly introduce high cardinality, such as email addresses or UUIDs, which can strain the database by creating excessive time series.

Because a time series is a unique combination of a metric name and its label pairs, the top 10 series count by metric names table can help you identify high-cardinality metrics. In case there's no single label that introduces performance issues, the combination of labels on a particular metric can still result in a high number of unique series. This table helps you pinpoint which metrics, through their label combinations, contribute most to high cardinality.

Furthermore, highlighting the top 10 label names with high memory usage helps identify labels that, although they may not be numerous, occupy substantial memory because of large or complex values.

Lastly, understanding the top 10 series count by label value pairs allows administrators to identify specific combinations that contribute heavily to the total series count.

The following table summarizes the different pieces of information available on the TSDB status page and their usage: